Electronic Structures of an [Fe(NNR
2
)]
+/0/-
Redox Series: Ligand
Noninnocence and Implications for Catalytic Nitrogen Fixation
Niklas B. Thompson
†
,
Paul H. Oyala
†
,
Hai T. Dong
‡
,
Matthew J. Chalkley
†
,
Jiyong Zhao
§
,
E.
Ercan Alp
§
,
Michael Hu
§
,
Nicolai Lehnert
*,‡
, and
Jonas C. Peters
*,†
†
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
California 91125, United States
‡
Department of Chemistry and Department of Biophysics, University of Michigan, Ann Arbor,
Michigan 48109-1055, United States
§
Advanced Photon Source (APS), Argonne National Laboratory (ANL), Argonne, Illinois 60439,
United States
Abstract
The intermediacy of metal-NNH
2
complexes has been implicated in the catalytic cycles of several
examples of transition-metal-mediated nitrogen (N
2
) fixation. In this context, we have shown that
triphosphine-supported Fe(N
2
) complexes can be reduced and protonated at the distal N atom to
yield Fe(NNH
2
) complexes over an array of charge and oxidation states. Upon exposure to further
H
+
/e
−
equivalents, these species either continue down a distal-type Chatt pathway to yield a
terminal iron(IV) nitride or instead follow a distal-to-alternating pathway resulting in N−H bond
formation at the proximal N atom. To understand the origin of this divergent selectivity, herein we
synthesize and elucidate the electronic structures of a redox series of Fe(NNMe
2
) complexes,
which serve as spectroscopic models for their reactive protonated congeners. Using a combination
of spectroscopies, in concert with density functional theory and correlated ab initio calculations,
we evidence one-electron redox noninnocence of the “NNMe
2
” moiety. Specifically, although two
closed-shell configurations of the “NNR
2
” ligand have been commonly considered in the literature
isodiazene and hydrazido(2−) we provide evidence suggesting that, in their reduced forms, the
present iron complexes are best viewed in terms of an open-shell [NNR
2
]
•-
ligand coupled
antiferromagnetically to the Fe center. This one-electron redox noninnocence resembles that of the
classically noninnocent ligand NO and may have mechanistic implications for selectivity in N
2
fixation activity.
*
Corresponding Authors: lehnertn@umich.edu, jpeters@caltech.edu.
Notes
The authors declare no competing financial interest.
Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website
at DOI:
10.1021/acs.inorgchem.9b00133
.
Additional synthetic details, experimental procedures, complete characterization data, and computational methods/models (
PDF
)
Accession Codes
CCDC
1890845
and
1890846
contain the supplementary crystallographic data for this paper. These data can be obtained free of charge
via
www.ccdc.cam.ac.uk/data_request/cif
, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge
Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033.
HHS Public Access
Author manuscript
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Published in final edited form as:
Inorg Chem
. 2019 March 04; 58(5): 3535–3549. doi:10.1021/acs.inorgchem.9b00133.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Graphical Abstract
1. INTRODUCTION
Since the pioneering studies of Chatt, Hidai, and co-workers,
1
the synthesis and reaction
chemistry of transition-metal complexes featuring terminal, doubly N,N-functionalized
dinitrogen (N
2
) ligands (“NNR
2
”) have been pursued because of the proposed intermediacy
of M(NNH
2
) species in thefixation of N
2
to ammonia (NH
3
).
2
In this context, the closed-
shell hydrazido(2−) configuration of the “NNR
2
” fragment is typically invoked to explain
the susceptibility of the distal N atom (N
β
) toward attack by electrophiles to produce metal
hydrazidium complexes, M(NNR
3
), en route to N−N bond cleavage.
3
At the same time,
many M(NNR
2
) complexes, especially those of the late transition metals, have been
characterized as adducts of the charge-neutral isodiazene (NNR) oxidation state.
2b
Valence
isomerization between these two closed-shell configurations−hydrazido(2−) and isodiazene
−has also been proposed.
4
Despite the prominence of Fe in the catalytic fixation of N
2
,
5
the corresponding chemistry of
Fe(NNR
2
) complexes is comparatively underdeveloped.
2b
,
6
Recently, we have characterized
[(P
3
B
)Fe(NNH
2
)]
+
[P
3
B
= tris(
o
-diisopropylphosphinophenyl)borane] as a plausible
intermediate in catalytic N
2
-to-NH
3
conversion by [(P
3
B
−3 3)Fe(N
2
)]
−
(Figure 1).
7
Upon
one-electron reduction to form the charge-neutral congener (P
3
B
)Fe(NNH
2
), this species can
be further protonated at N
β
to yield NH
3
and a terminal iron(IV) nitride, [(P
3
B
)Fe
≡
N]
+,
8
consistent with hydrazido(2−)-like reactivity. In comparison, the isoelectronic and
isostructural complex [(P
3
Si
)Fe(NNH
2
)]
+
[P
Si
3
= tris(
o
-
diisopropylphosphinophenyl)silylide] appears to be stable toward protonation at low
temperature.
9
Instead, this species can be further reduced to the formally 19e
−
complex
(P
3
Si
)Fe(NNH
2
), which disproportionates to produce complex mixtures that, most notably,
include the hydrazine adduct [(P
3
Si
)Fe(N
2
H
4
)]. When [(P
3
Si
)Fe(NNH
2
)]
+
is formed in situ
and subsequently treated with substoichiometric CoCp*
2
(Cp* =
pentamethylcyclopentadienide), high yields of both [(P
3
Si
)-Fe(N
2
H
4
)]
+
and (P
3
Si
)Fe(N
2
)
are produced (Figure 1), representing, on balance, the exchange of two H atom equivalents
between the [(P
3
Si
)Fe(NNH
2
)]
+/0
redox pair to effect functionalization of the proximal N
atom (N
α
).
9
Thompson et al.
Page 2
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Preliminary electron paramagnetic resonance (EPR) data of (P
3
Si
)Fe(NNH
2
) and its
alkylated analogue (P
3
Si
)Fe(NNMe
2
) reveal significant spin density on at least a single N
atom (presumably N
α
), which may serve a functional role in promoting proton-coupled
electron transfer (PCET) chemistry at this center.
10
Although [(P
3
Si
)Fe(N)]
−
2 is a
comparatively inefficient catalyst for N
2
fixation,
11
we have recently proposed that similar
PCET processes may play a role in increasing the efficiency of N
2
fixation by [(P
3
B
)Fe(N
2
)]
−
when using metallocene-based reductants.
12
Given this behavior, we were curious whether
the electronic structures of [(P
3
E
)Fe(NNR
2
)]
n
complexes might feature a significant weight
of the intermediate, open-shell hydrazyl radical anion configuration ([NNR
2
]
•−
) of this
redox-active ligand. Although this redox state has not, to our knowledge, been explored
experimentally,
13
such a proposal is reasonable given the low-energy
π
* orbital of both
parent and N,N-dialkylisodiazenes,
14
which thus bear a resemblance to the classically redox-
noninnocent nitrosyl ligand (NO).
15
Indeed, both neutral and cationic hydrazyl radicals
([HNNR
2
]
•
/ [H
2
NNR
2
]
•+
; R = H, alkyl) have been characterized in their free forms.
16
Such
an electronic structure would be conceptually similar to that proposed for metal imidyl
([NR]
•−
) and aminyl ([NR
2
]
•
) complexes that promote PCET reactivity.
10
A detailed characterization of the electronic structures of [(P
3
E
)Fe(NNH
2
)]
n
is hampered by
their high reactivity.
7
–
9
However, previous work has shown that the N,N-dimethylated
complexes [(P
3
E
)Fe(NNMe
2
)]
n
are excellent spectroscopic models for their protonated
congeners, while exhibiting greater stability.
8
,
9
Herein, we exploit this stability to
characterize the nature of the Fe−NNMe
2
interaction across the [(P
3
B
)Fe-(NNMe
2
)]
+/0/−
redox series. This rich redox chemistry allows us to study representative complexes that are
isoelectronic to each [(P
3
B
)Fe(NNH
2
)]
n
and [(P
3
Si
)Fe(NNH
2
)]
n
species implicated thus far
in N
2
fixation chemistry mediated by the (P
3
E
)Fe platform. Through a combination of
spectroscopic and computational techniques, we demonstrate that these “hydra-zido”
complexes possess ground and low-lying excited states that result from antiferromagnetic
exchange coupling between the Fe and an open-shell [NNR
2
]
•-
ligand. A discussion of the
relevance of these electronic structures to the N
2
fixation activity of this class of complexes
is presented.
2. RESULTS
2.1. Synthesis and Structural Analysis of [(P
3
B
)Fe-(NNMe
2
)]
+/0/-
.
Like its protonated analogue, (P
B
3
)Fe(NNMe
2
) has a diamagnetic ground state; the cyclic
voltammogram (CV) of (P
3
B
)Fe(NNMe
2
) shows a reversible oxidation centered at −1.16 V
versus Fc
+/0
in tetrahydrofuran (THF).
8
Accordingly, its one-electron oxidation by [FeCp*
2
]
[BAr
F
4
] ([BAr
F
4
]
−
= tetrakis[3,5-bis(trifluoromethyl)phenyl]borate) yields [(P
3
B
)-
Fe(NNMe
2
)][BAr
F
4
], which populates an S = ½ ground state (Figure 2A).
[(P
3
B
)Fe(NNMe
2
)]
+
is moderately stable in the solid state and in solution but decomposes to
an intractable mixture upon prolonged heating at 70 °C.
The CV of (P
3
B
)Fe(NNMe
2
) also shows a quasi-reversible reduction centered around −2.65
V versus Fc
+/0
in THF. This couple becomes increasingly reversible at higher scan rates
(Figure S34), prompting us to explore whether the reduction product could be characterized
in situ at low temperature. Indeed, the reduction of (P
3
B
)Fe(NNMe
2
) in 2-
Thompson et al.
Page 3
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
methyltetrahydrofuran (2-MeTHF) with stoichiometric KC
8
at −78 °C produces an
S
= ½
species that, on the basis of its distinctive Mössbauer and EPR properties (vide infra), can be
assigned as [(P
3
B
)Fe(NNMe
2
)]
−
(Figure 2A). [(P
3
B
)Fe(NNMe
2
)]
−
isunstable, decomposing
within minutes upon warming to ambient temperatures. Nevertheless, we were able to obtain
a single crystal of the [K(benzo-15-crown-5)
2
]
+
salt of [(P
3
B
)Fe-(NNMe
2
)]
−
suitable for X-
ray diffraction [see the Supporting Information (SI) for details], confirming the structural
assignment made on the basis of spectroscopy.
The solid-state structures of [(P
3
B
)Fe(NNMe
2
)]
+/0/-
determined by X-ray diffraction are
shown in Figure 2B. For each complex, the “NNMe
2
” ligand is planar [∑
∠
(N
β
) = 360°], and
an approximate mirror plane is formed by P1, Fe, and N
α
. The equatorial phosphine
substituents of [(P
3
B
)Fe(NNMe
2
)]
0/-
adopt a slightly distorted trigonal arrangement about
the Fe center with geometries intermediate between trigonal-bipyramidal and tetrahedral (
τ
= 0.34 and 0.53, respectively).
17
Upon oxidation to [(P
3
B
)Fe(NNMe
2
)]
+
, the P2−Fe−P3
angle widens as an
η
3
-
B
,
C
,
C
interaction forms between the Fe center and the phenylene
linker of one of the phosphine substituents [
d
(Fe− C1) = 2.727(3) Å;
d
(Fe−C2) = 2.683(3)
Å] (see Figure 2B). While the Fe−N−N angle is close to linear for the charge-neutral
complex (176.1°), it becomes significantly bent in [(P
B
3
)Fe(NNMe
2
)]
+
(159.6°) and
[(P
3
B
)Fe(NNMe
2
)]
−
(161.7°). Similar Fe−N−N angles were observed for [(P
3
B
)Fe-(NNH
2
)]
+
(~150°)
7
and (P
Si
3
)Fe(NNH
2
) (~151°),
9
suggesting that these alkylated complexes are
faithful structural models of their protonated analogues.
In this redox series, the Fe−N/N−N distances change in a nonlinear fashion, from 1.738(3)/
1.252(4) Å in [(P
3
B
)Fe-(NNMe
2
)]
+
to 1.680(2)/1.293(3) Å in (P
3
B
)Fe(NNMe
2
) to1.771(7)/
1.27(1) Å in [(P
3
B
)Fe(NNMe
2
)]
−
. This range of N−N distances is longer than that calculated
for free NNR
2
(1.20−1.22 Å
14c
,
18
) but significantly shorter than that observed for free
hydrazine (1.47 Å
19
) or M(NNR
3
) (1.40−1.43 Å; R = H, alkyl
3e
,
20
) complexes, suggesting
some degree of N−N
π
bonding. The Fe−N distance of (P
3
B
)Fe(NNMe
2
) falls within the
range observed for terminal iron imido complexes (1.61−1.72 Å
21
) but is longer than those
observed for C
3
-symmetric, terminal iron nitrides (1.51−1.55 Å
8
,
22
). The Fe−N distances of
[(P
B
)Fe(NNMe
2
)]
+/−
are longer than those typically observed for iron imides but do fall into
this range if the four-coordinate iron(III) imidyl species reported by Betley and co-workers
are included (Fe−N = 1.77 Å).
10e
,
f
2.2.
57
Fe Mössbauer Spectroscopy.
The Mössbauer spectrum of (P
3
B
)Fe(NNMe
2
) has been reported in a recent communication,
8
and its parameters are collected in Table 1, along with those of related (P
3
E
)Fe complexes.
While (P
3
B
)Fe-(NNMe
2
) has parameters very similar to those of (P
3
B
)Fe-(NNH
2
) and the
isostructural, silylated complex (P
3
B
)Fe(NN-[Si
2
]) (N[Si
2
] = 2,2,5,5-tetramethyl-1-aza-2,5-
disilacyclopentyl), the isomer shift of the nominally isoelectronic terminal imido complex
(P
3
B
)Fe(NAd) is significantly smaller than those of the (P
3
B
)Fe(NNR
2
) complexes. This
difference can be attributed to greater covalency in the Fe−N interaction of the imido
complex,
8
,
11
,
23
which is consistent with the structural analysis given above. The Mössbauer
parameters of the isoelectronic P
3
Si
complexes [(P
3
Si
)Fe(NNR
2
)]
+
(R = H, Me) are very
similar to those of their P
3
B
analogues, despite the difference in the axial ligating elements.
Thompson et al.
Page 4
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
For the [(P
3
B
)Fe(NNMe −)]
+/0/-
series, the isomer shift of the cationic complex increases by
0.14 mm s
−1
relative to its charge-neutral congener, while that of the anionic complex
increases by0.22 mm s
−1
, reflecting the increased Fe−N and Fe−Pdistances observed
crystallographically. As above, the Mössbauer parameters of [(P
3
B
)Fe(NNMe
2
)]
+
are close
to those reported for its protonated analogue,
7
further illustrating the utility of the alkylated
complexes as spectroscopic models. Interestingly, even at temperatures as high as 80 K,
[(P
3
B
)Fe(NNMe
2
)]
+
is in the limit of slow electronic relaxation, which has allowed us to
estimate the
57
Fe hyperfine coupling (HFC) tensor from the field dependence of the
spectrum (Table 1 and Figure 3A).
[(P
3
B
)Fe(NNMe
2
)]
−
also exhibits unusually slow electronic relaxation at 80 K (Figure 3B),
which was observed for the isoelectronic silyl complexes (P
3
Si
)Fe(NNR
2
) (R = H, Me).
9
The
similarity between the Mössbauer parameters of (P
3
Si
)Fe-(NNR
2
) (R = H, Me) and those of
[(P
3
B
)Fe(NNMe
2
)]
−
(Table 1) reveals that this set of isoelectronic complexes exhibit similar
electronic structures, as was observed for their one-electron-oxidized congeners above.
Notably, the isotropic
57
Fe HFC constant (
a
iso
) of [(P
3
B
)Fe(NNMe
2
)]
−
is more than 2 times
larger than that of [(P
3
B
)Fe(NNMe
2
)]
+
; indeed, a similar trend in
a
iso
is observed for nearly
every magnetic nucleus in the coordination sphere of Fe (vide infra). The Mössbauer
parameters calculated for [(P
3
B
)Fe(NNMe
2
)]
+/−
via density functional theory (DFT) are in
excellent agreement with the experimental ones extracted from simulation (Table 1).
2.3. Nuclear Resonance Vibrational Spectroscopy (NRVS) for the [(P
3
B
)Fe(NNMe
2
)]
+/0/-
Series.
To analyze in detail the changes that occur in the bonding of the Fe−NNMe
2
unit upon
oxidation/reduction, we collected NRVS data for the full redox series, including the [(P
3
B
)
57
Fe(
14
N
14
NMe
2
)]
n
and [(P
3
B
)
57
Fe(
15
N
15
NMe
2
)]
n
isotopologues. This vibrational technique
selectively probes motions coupled to the
57
Fe nucleus and, with the aid of isotopic labeling
and quantum-chemistry-centered normal coordinate analysis (QCC-NCA),
24
allows us to
extract force constants for the metal−ligand stretching and bending modes.
The experimental NRVS data of neutral (P
3
B
)Fe(NNMe
2
) show two major isotope-sensitive
features at 448 and 545 cm
−1
that shift to 442 and 540 cm
−1
, respectively, upon
15
N labeling.
DFT calculations reproduce the NRVS data of each complex studied here particularly well
and, therefore, form a reliable basis for further refinement using QCC-NCA. Figure 4A
shows a comparison of the experimental and QCC-NCA spectra for (P
3
B
)Fe(NNMe
2
). On
the basis of our analysis, the band at 545 cm
−1
corresponds to the Fe−N stretch, whereas the
448 cm
−1
feature is the Fe−N−N out-of-plane bend. Our simulations further show that the
weak features at higher energy than the Fe−N stretch, which are obscured by the background
noise but clearly visible in the QCC-NCA simulation at 572 and 605 cm
−1
, have Fe−N−N
in-plane bending character. Note that, in the QCC-NCA simulations, the Fe−N−N in plane
bend is distributed over multiple features in the 570−610 cm
−1
region. Labeling of the N
−CH groups with
13
C reveals three isotope-sensitive features in the IR spectrum of the
neutral complex, at 1337, 1146, and 874 cm
−1
(Figure S28). These same features shift upon
15
N labeling, demonstrating significant mode mixing between N−N and N−C vibrations,
although the QCC-NCA analysis suggests that the mode at 1337 cm
−1
is predominantly N
Thompson et al.
Page 5
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
−N stretching, while the two lower-energy modes are dominated by the antisymmetric and
symmetric N−C stretching modes, respectively. Table S24 summarizes our assignments.
In the NRVS data of [(P
3
B
)Fe(NNMe
2
)]
−
, two isotope-sensitive bands are observed at 428
and 541 cm
−1
, which shift to 422 and 530 cm
−1
with
15
N labeling, respectively. The feature
at 541 cm
−1
belongs to a mixed Fe−N stretch and Fe−N−N in-plane bending mode, where
the strong mixing between these two internal coordinates is facilitated by the now bent
structure of the Fe−N−N unit. The feature at 428 cm
−1
is assigned to the Fe−N−N out-of-
plane bend based on QCC-NCA. Interestingly, the intense feature at 346 cm
−1
in the NRVS
data corresponds to a mixed N−N−C/C−N−C bend. This mode also has distinct Fe−N
stretching character (19% contribution). However, in this normal mode, the two N atoms do
not move, and, hence, this feature is, in fact, not
15
N isotope-sensitive. Figure 4B shows a
comparison of the experimental and QCC NCA spectra. Owing to its thermal sensitivity, IR
spectra of [(P
3
B
)Fe(NNMe
2
)]
−
were not collected.
In the cationic compound [(P
3
B
)Fe(NNMe
2
)]
+
, the isotope-sensitive bands are observed at
588 and 495 cm
−1
, which shift to 579 and 486 cm
−1
upon
15
N substitution, respectively.
Here, the 588 cm
−1
feature corresponds to the in-plane bend, mixed with the Fe−N stretch.
The assignment of the 495 cm
−1
band is less clear because both the Fe−N stretching and Fe
−N−N out-of-plane bending modes occur in this energy region. The QCCNCA simulations
predict that 66% of the NRVS vibrational density of states (VDOS) intensity of the 495 cm
−1
band originates from the Fe−N stretch. The feature around 348 cm
−1
is similar in nature
to that described for the anionic compound and shows distinct Fe−N stretching character.
Figure 4C compares the experimental and QCC-NCA spectra, showing again excellent
agreement between the simulations and experiment. IR spectra of [(P
3
B
)Fe(NNMe
2
)]
+
show
a
15
N-sensitive feature at 1495 cm
−1
and another band at 1371 cm
−1
, the latter of which also
shifts upon
13
C labeling of the N−CH
3
position (Figure S29). DFT calculations predict
another feature at 1153 cm
−1
, sensitive to both
15
N and
13
C labeling, which, however, is
obscured by intense resonances from the [BAr
F
4
]
−
counterion in the experimental spectra.
On the basis of our analysis (Table S24), the mode at 1495 cm
−1
corresponds mostly to N−N
stretching, while the lower-energy modes are mixtures of N−N and antisymmetric N−C
stretching.
Table S24 summarizes all of our assignments and compares the experimentally observed
vibrational frequencies to those obtained from the QCC-NCA simulations. As is evident
from the table, a very intense degree of mode mixing within the Fe− NNMe
2
unit is present
in all three complexes, especially the cationic and anionic compounds with the bent Fe−N
−N units. It is therefore not possible to draw conclusions about changes in bonding between
the three complexes based solely on the vibrational energies. As discussed below in section
3.1, such conclusions rely instead on the force constants from the QCCNCA simulations,
listed in Table 2 (see also Table S25).
2.4. Low-Lying Excited States of [(P
3
B
)Fe(NNMe
2
)]
+/0
.
Although (P
3
B
)Fe(NNMe
2
) is a diamagnet in its ground state, preliminary variable-
temperature (VT) NMR and DFT studies have evidenced the presence of a low-lying,
S
= 1
paramagnetic excited state.
8
Similar behavior was observed for the isoelectronic complex
Thompson et al.
Page 6
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
[(P
3
Si
)Fe(NNMe
2
)]
+
9
. In both of these cases, fitting the VT NMR data to a simple two-state,
Boltzmann-weighted magnetization function showed that these triplet states lie only 3.7
± 0.1 and 6.7 ± 0.3 kcal mol
−1
(1300 ± 30 and 2300 ± 100 cm
−1
) above the diamagnetic
ground states, respectively.
8
,
9
It is noteworthy that the entropic contributions to these energy
differences appear to be small, and we have obtained a more precise estimate of the
adiabatic singlet−triplet gapof (P
3
B
)Fe(NNMe
2
) of 1266 ± 7 cm
−1
, assuming ΔG
≈
ΔH (see
the SI for details).
VT
15
N NMR studies of (P
3
B
)Fe(
15
N
15
NMe
2
) suggest that both N
α
and N
β
accumulate spin
density in this triplet excited state (Figure 5A). Moreover, an examination of the VT
1
H
NMR data shows that the isotropic shift of the N−CH
3
protons is roughly an order of
magnitude greater than that of any of the protons on the P
3
B
ligand in the excited state
(Figure 5B). For example, taking the methine protons of the isopropyl substituents on the
phosphines as a reference, the influence of the Fe ion in the excited state can be quantified
by a Curie factor of approximately 1 × 10
5
ppm K. The Curie factor of the N− CH
3
protons
is approximately 10 × 10
5
ppm K, despite the fact that these protons are separated from the
Fe ion by an additional bond. This striking observation suggests that significant spin density
is localized on the nitrogenous ligand in the triplet excited state of (P
3
B
)Fe(NNMe
2
).
Given the behavior of (P
3
B
)Fe(NNMe
2
), we were curious whether [(P
3
B
)Fe(NNMe
2
)]
+
also
populates an excited state of higher multiplicity at relatively low temperatures. Solution-
phase VT magnetic susceptibility measurements are consistent with this hypothesis,
revealing an increase of almost 1
β
e
as the temperature is raised over a 140 °C range. While
it is not possible to determine the excited-state multiplicity from these data alone, DFT
calculations indicate that the first sextet state is significantly higher in energy than the first
quartet state (see the SI). Using a two-state model similar to that above, we estimate that the
quartet state lies only ~5 kcal mol
−1
(~1700 cm
−1
) above the doublet ground state of
[(P
3
B
)Fe(NNMe
2
)]
+
from these susceptibility measurements.
2.5. EPR Studies of [(P
3
B
)Fe(NNMe
2
)]
+/−
.
The indirect evidence for relatively large ligand-centered spin density in the triplet excited
state of (P
3
B
)Fe(NNMe
2
) motivated us to experimentally determine the spin-density
distribution within the “NNMe
2
” ligands of [(P
3
B
)Fe(NNMe
2
)]
+/−
using EPR-based
techniques. With regard to the nature of the Fe−NNMe
2
interaction, these data complement
the structural and vibrational data presented thus far because the complete anisotropic HFC
tensor is a sensitive probe of the valence electronic structure of a particular nucleus.
The continuous-wave (CW) X-band EPR spectrum of [(P
3
B
)Fe(NNMe
2
)]
+
is shown in
Figure 6, along with that of [(P
3
B
)Fe(NNH
2
)]
+
, for comparison. The overall rhombicity and
anisotropy of the g tensors of [(P
B
3
)Fe(NNMe
2
)]
+
closely matches that of the protonated
complex [(P
B
3
)Fe(NNH
2
)]
+
(Table 3),
7
once again validating the use of the former as a
spectroscopic model of the latter. This is in contrast to the g tensors of the nominally
isoelectronic imido species, [(P
3
B
)Fe-(NAd)]
+
, which is significantly more anisotropic
(Table 3).
Thompson et al.
Page 7
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
The CW spectrum of [(P
3
B
)Fe(NNMe
2
)]
+
shows clear HFC to three
31
Pnuclei at g
1
= 2.005.
In order to measure the couplings to the rest of the magnetic nuclei in [(P
3
B
)Fe-(NNMe
2
)]
+
,
we turned to pulse experiments at the Q band (~34 GHz). The electron nuclear double
resonance (ENDOR) spectra (Figure S12) reveal coupling to
11
B and three
31
P nuclei, as
well as to two nearly equivalent
13
C nuclei in [(P
3
B
)Fe(NN-(
13
CH
3
)
2
)]
+
. The HFCs to N
α
/
N
β
are quite anisotropic and were measured via hyperfine sublevel correlation (HYSCORE)
spectroscopy using
14
N and
15
N isotopologues (Figures S20−S27). The complete set of
ligand HFC tensors is given in Table 4. We assign the more strongly coupled N atom as N
α
,
and note that the
14
N
α
coupling at
g
2
(2.089) of A
2
= −5.7 MHz is quite similar to that
measured previously for [(P
3
B
)Fe(NNH
2
)]
+
at
g
2
(2.091), A
2
= −6.4 MHz;
7
the sign of the
coupling is assumed here to be negative to be consistent with the published ENDOR data for
[(P
3
B
)Fe(NNH
2
)]
+
. Because of the anisotropy of the hyperfine interactions, these ENDOR
studies were unable to unambiguously define the complete HFC tensor for N
α
of
[(P
3
B
)Fe(NNH
2
)]
+
, consistent with the extreme rhombicity of the analogous tensor
confirmed here for [(P
3
B
)Fe(NNMe
2
)]
+
via HYSCORE spectroscopy. Similar rhombicity is
also observed for the 14N
α
couplings of the anionic complex, [(P
3
B
)Fe(NNMe
2
)]
−
(Table 4).
[(P
3
B
)Fe(NNMe
2
)]
−
possesses low
g
anisotropy, with a
g
tensor that is quite similar to those
of the isoelectronic silyl complexes, (P
3
Si
)Fe(NNR
2
) (R = H, Me; Table 3), reinforcing the
idea that these species exhibit similar electronic structures (vide infra). Parts A and B of
Figure 7 show the second-derivative CW X-band EPR spectra of [(P
3
B
)Fe(
14
N
14
NMe
2
)]
−
and [(P
3
B
)Fe(
15
N
15
NMe
2
)]
−
, along with simulations. Although these simulations contain six
independent HFC tensors, those of the 14/15N
α
, 14/15N
β
, 11B, and 31P
γ
nuclei were
determined independently via Q-band ENDOR and HYSCORE spectroscopy, as above
(Figures S8 and S14−S19). The HFC tensors of the remaining two, more strongly coupled,
31
P nuclei (
31
P
α
/
β
) were determined through simultaneous fitting of the X-band CW and
ENDOR data (Figure S9). As can be seen from the
14
N−
15
N difference spectrum shown in
Figure 7C, the final simulation is of high quality. The complete set of ligand HFCs is given
in Table 4.
Q-band ENDOR spectroscopy resolves a single
13
C HFC tensor for [(P
3
B
)Fe(NN(
13
CH
3
)
2
)]
−
, whose isotropic component is more than twice the magnitude of the average
13
C coupling
observed for [(P
3
B
)Fe(NN(
13
CH
3
)
2
)]
+
(Table 4). Considering that the
13
C nuclei are >3.8 Å
from the Fe center, the magnitude of this coupling is surprisingly large, corresponding to a C
2s spin population of ~0.5%; for comparison, the most strongly coupled
31
P nucleus, which
is bound directly to the Fe center, has a P 3s spin population of ~0.9%.
25
This observation is
reminiscent of the VT
1
H NMR data presented above for (P
3
B
)Fe(NNMe
2
). An examination
of Table 4 shows that the magnitude of the isotropic ligand couplings almost uniformly
increases upon two-electron reduction of [(P
3
B
)Fe(NNMe
2
)]
+
to form [(P
3
B
)Fe(NNMe
2
)]
−
.
The only exceptions are a single
31
P nucleus, which remains almost unchanged, and N
β
,
which has a decreased isotropic component (although the anisotropic coupling increases,
vide infra). This pattern matches the observation by Mössbauer spectroscopy of nearly 2-
fold-increased
57
Fe HFC upon reduction (vide supra).
Thompson et al.
Page 8
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
2.6. Quantum-Chemical Calculations.
2.6.1. Preliminary Bonding Considerations.—
As a free molecule, NNH
2
exhibits a
planar, C
2
v
geometry.
14c
The important orbitals for interaction with a transition-metal center
consist of an a
1
-symmetry N
α
lone pair (
σ
N
), a b
1
-symmetry N
α
lone pair (
π
N
), and the
orthogonal b
2
-symmetry
π
* orbital (
π
*
NN
). The qualitative molecular orbital (MO) picture
for a charge-neutral, linear R
′
3
B−Fe−NNR
2
fragment under pseudo-
C
2
v
symmetry is shown
in Figure 8A.
In the limit of large orbital overlap, mixing of
π
N
and
π
*
NN
with the 3d
xz
and 3d
yz
orbitals,
respectively, should result in the formation of two
π
bonds, one “in-plane” (b
1
symmetry)
and one “out-of-plane” (b
2
symmetry). This simple imide-like bonding situation could be
represented by an Fe
≡
N−NR
2
valence bond (VB) picture;
26
however, if N
β
donates its lone
pair to the b
2
-symmetry
π
bond, this will lift the degeneracy of the two orthogonal
π
interactions (denoted by
Δ
ε
π
in Figure 8A) and produce a frontier
π
-orbital system isolobal
to ketene,
27
i.e., an Fe
≡
N−NR
2
VB picture. Bending of the Fe−N−NR
2
angle, as observed in
[(P
3
B
)Fe(NNMe
2
)]
+/−
, is expected to result in rehybridization of the
π
N
lone pair, producing
VB pictures corresponding to the bent imido (Fe
≡
N−NR
2
) and isodiazene adduct (Fe
≡
N
−NR
2
) structures shown in Figure 8B.
While these VB representations apply to the case of purely covalent bond character, in the
limit of small orbital overlap, the Fe−NNR
2
out-of-plane
π
bonding may be better described
in terms of an exchange coupling interaction involving anti-ferromagnetic alignment of the
π
-symmetry electrons, that is, a formal [NNR
2
]
•−
configuration for the ligand. In this limit,
the bonding interaction within the Fe−NNR
2
moiety would take on diradical character.
28
2.6.2. DFT Calculations.—
To help delineate the relative weights that the VB
representations shown in Figure 8B contribute to the electronic structures of [(P
3
B
)Fe-
(NNMe
2
)]
+/0/-
, we first performed a series of DFT calculations using the TPSSh functional.
This approach was previously shown to be accurate in predicting Mössbauer and X-ray
adsorption spectroscopy (XAS) spectra for (P
3
B
)Fe complexes (e.g., Table 1).
8
Starting with
[(P
3
B
)Fe(NNMe
2
)]
−
, the optimized Kohn−Sham wave function is significantly spin-
contaminated (
〈
Ŝ
2
〉
= 1.02), which is an indicator of broken-symmetry (BS) character.
Indeed, an examination of the unrestricted corresponding orbitals from this calculation
reveals a pair of magnetic orbitals corresponding to antiferromagnetic coupling of majority
spin 3d
yz
and minority spin
π
*
NN
electrons. The overlap between these two spin orbitals
(
〈
α
|
β
〉
= 0.89) is in the range typically taken to correspond to metal−ligand
antiferromagnetic coupling in, for example, Fe(NO) complexes,
29
thus arguing in favor of a
[NNMe
2
]
•-
configuration for the “hydrazido” ligand. In accordance with the qualitative MO
diagram of Figure 8A, the uncoupled metal-based singly occupied molecular orbital
(SOMO) is of 3d
xz
character (Figure 9A).
At this level of theory, only a closed-shell singlet configuration could be converged for
(P
3
B
)Fe(NNMe
2
) in its ground state. Similarly, an unrestricted calculation of
[(P
3
B
)Fe(NNMe
2
)]
+
is relatively spin-uncontaminated, corresponding to a normal two-
electron 3d
yz
/
π
*
NN
interaction (27E8
〈
α
|
β
〉
= 0.97). However, given that BS DFT
calculations are highly sensitive to the percentage of Hartree−Fock exchange incorporated in
Thompson et al.
Page 9
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
the exchange-correlation functional, these results are ambiguous. For example, increasing
the Hartree−Fock exchange from 10% to 25% (a value closer to that typically used in BS-
DFT studies
29
) results in BS (antiferromagnetic) wave functions for both (P
3
B
)Fe-(NNMe
2
)
and [(P
3
B
)Fe(NNMe
2
)]
+
in their ground spin states (
〈
α
|
β
〉
= 0.85 and 0.90 for the 3d
yz
/
π
*
NN
interaction, respectively). Using 25% exact exchange in the calculation decreases the
corresponding overlap integral to 0.70 for [(P
3
B
)Fe(NNMe
2
)]
−
.
〉
Repeating these calculations for the low-lying triplet and quartet excited states of
(P
3
B
)Fe(NNMe
2
) and [(P
3
B
)Fe-(NNMe
2
)]
+
, respectively, produces BS solutions similar to
that of [(P
3
B
)Fe(NNMe
2
)]
−
(Figure 9B,C). The antiferromagnetic nature of the out-of-plane
Fe−NNMe
2
π
bonding persists, although in the case of [(P
3
B
)Fe(NNMe
2
)]
+
, the strength of
the magnetic coupling increases, reflected by the enhanced orbital overlap (
〈
α
|
β
〉
= 0.94).
30
In the optimized structures of these redox and spin states, the Fe−N−N angle is significantly
bent (156−164°), which can be rationalized by population of the 3d
xz
orbital, which would
result in a strongly antibonding in-plane
π
interaction with the
π
N
electrons in a linear
geometry. Once more, the strength of the magnetic coupling between the 3d
yz
/
π
*
NN
electrons is highly sensitive to the amount of exact exchange included in the calculation. The
corresponding overlapintegrals decrease to 0.74 and 0.89 for (P
3
B
)Fe(NNMe
2
) (
S
= 1) and
[(P
B
3
)Fe(NNMe)]
+
(
S
= 3/
2
), respectively, upon an increase in the the percentage of Hartree
−Fock from 10% to 25%.
2.6.3. Complete-Active-Space Self-Consistent-Field (CASSCF) Calculations.
—
To resolve this ambiguity, we turned to wave-function-based calculations. In contrast to
approximate, single-reference methods like BS DFT, multireference techniques such as the
CASSCF method can properly represent open-shell singlet wave functions, as well as more
general cases of metal−ligand exchange coupling.
From a ground-state-specific calculation of (P
3
B
)Fe-(NNMe
2
), some multireference
character is evident, with the closed-shell configuration
3d
xy
2
3d
x
2
−
y
2
2
3d
xz
+
π
N
2
3d
z
2
+2p
z
2
3d
yz
+
π
NN
*
2
composing 79.8% of the zeroth-order wave function. The antiferromagnetic nature of the
remaining determinants is hinted at from localization of the active space orbitals, which
results in strong spatial separation of the Fe 3d
yz
and
π
*
NN
orbitals (Figure 10).
28c
,
d
A
quantitative measure of the antiferromagnetic character of the 3d
yz
/
π
*
NN
interaction can be
obtained from the occupation numbers of the corresponding bonding and antibonding
natural orbitals (
n
±
). Table 5 compiles an index of the diradical character (Y) of these
interactions, with
Y
= 0 corresponding to a pure two-electron bond and
Y
= 1 corresponding
to a pure diradical.
28b
In this case, the significant diradical character of 3d
yz
/
π
*
NN
interaction (17%) is consistent with the presence of a spin-coupled [NNMe
2
]
•-
ligand in the
singlet ground state of (P
3
B
)Fe(NNMe
2
). Comparable diradical character (20−30%) is found
from in the
π
bonding of porphyrin-supported Fe(NO) complexes.
28c
Thompson et al.
Page 10
Inorg Chem
. Author manuscript; available in PMC 2020 March 04.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript